Cyber Threat Brief — Issue #011

On June 11, 2026, a threat group called Icarus compromised Klue, an AI-powered competitive intelligence platform used by sales and marketing teams to track competitor activity. Klue integrates deeply with enterprise CRM systems including Salesforce, Gong, HubSpot, SharePoint, and Slack. The attack started with a legacy credential from a discontinued integration project that nobody had revoked. Using that credential, Icarus pushed a code update to Klue's backend that harvested OAuth tokens for every connected customer. Those tokens gave the attackers direct access to each customer's Salesforce environment without needing a password. The resulting victim list reads like an industry directory. LastPass, HackerOne, Huntress, Recorded Future, Tanium, Jamf, Snyk, OneTrust, Sprout Social, BeyondTrust, 8x8, and Pendo have all confirmed impact. More are expected to come forward. The data stolen was primarily business contact information and CRM records, including names, email addresses, phone numbers, physical addresses, and customer support case data. No core product infrastructure or customer secrets were affected at any of the named organizations. Icarus began posting stolen data on its leak site on June 22 and has been sending extortion demands to affected companies. Salesforce disabled the Klue Battlecards integration on June 17 in response, closing the exposure window that ran from June 11 through June 17.

Security researchers at SentinelOne discovered a new North Korean macOS malware family this week called Gaslight, a Rust-based backdoor and information stealer. The malware establishes persistence on the machine, connects to attacker infrastructure through Telegram, steals browser passwords, macOS keychain data, and system information, and uploads everything in a compressed file. None of that is new. What is new is what Gaslight does to the security tools that try to analyze it. The binary contains a 3.5 kilobyte block of 38 fabricated error messages designed to look like internal system failures inside an AI analysis tool. When a security analyst uses an AI assistant to examine the malware, these fake messages flood the AI's context — fake memory errors, fake token expiry warnings, fake disk failures, fake analysis flags — all designed to make the AI doubt its own session and stop the investigation before it completes. The technique did not successfully bypass any production AI analysis platform in current testing. But earlier North Korean macOS malware used a single injected block for the same purpose. Gaslight uses 38. Someone is counting failures and improving the approach. The direction of travel is clear.

New reporting this week confirmed that the FortiBleed credential harvesting campaign is larger than previously disclosed. The operation, now attributed by multiple intelligence firms to a Russian-speaking initial access broker, has been active since February 2026 and has targeted more than 430,000 FortiGate firewall devices globally. The scale means this is not a targeted campaign. It is a systematic sweep of every exposed Fortinet VPN device on the internet. Compromised credentials are being packaged and sold to other threat actors through criminal markets, meaning the organizations affected are not just at risk from the FortiBleed operators. They are at risk from whoever buys the access. If your organization has an internet-facing Fortinet VPN appliance and has not changed its credentials since February, that access is likely already for sale.